Speciation of ambient fine organic carbon particles and source

apportionment of PM2.5 in Indian cities

Zohir Chowdhury,1,2 Mei Zheng,1 James J. Schauer,3 Rebecca J. Sheesley,3

Lynn G. Salmon,4 Glen R. Cass,1,5,6 and Armistead G. Russell5

Received 2 January 2007; revised 28 March 2007; accepted 16 April 2007; published 7 August 2007.

[1] Fine particle organic carbon in Delhi, Mumbai, Kolkata, and Chandigarh is speciatedto quantify sources contributing to fine particle pollution. Gas chromatography/massspectrometry of 29 particle-phase organic compounds, including n-alkanes, polycyclicaromatic hydrocarbons (PAHs), hopanes, steranes, and levoglucosan along withquantification of silicon, aluminum, and elemental carbon are used in a molecular-markerbased source apportionment model to quantify the primary source contributions to thePM2.5 mass concentrations for four seasons in three sites and for the summer inChandigarh. Five primary sources are identified and quantified: diesel engine exhaust,gasoline engine exhaust, road dust, coal combustion, and biomass combustion. Importanttrends in the seasonal and spatial patterns of the impact of these five sources areobserved. On average, primary emissions from fossil fuel combustion (coal, diesel, andgasoline) are responsible for about 25–33% of PM2.5 mass in Delhi, 21–36% in Mumbai,37–57% in Kolkata, and 28% in Chandigarh. These figures can be compared to thebiomass combustion contributions to ambient PM2.5 of 7–20% for Delhi, 7–20% forMumbai, 13–18% for Kolkata, and 8% for Chandigarh. These measurements provideimportant information about the seasonal and spatial distribution of fine particle phaseorganic compounds in Indian cities as well as quantifying source contributions leading tothe fine particle air pollution in those cities.

Citation: Chowdhury, Z., M. Zheng, J. J. Schauer, R. J. Sheesley, L. G. Salmon, G. R. Cass, and A. G. Russell (2007), Speciation of

ambient fine organic carbon particles and source apportionment of PM2.5 in Indian cities, J. Geophys. Res., 112, D15303, doi:10.1029/

2007JD008386.

1. Introduction

[2] Urban areas in India experience very high concen-trations of airborne fine particulate matter (PM) (i.e., PMwith an aerodynamic diameter less than 2.5 mm, PMFine, orPM2.5), [United Nations Development Programme andWorld Bank Energy Sector Management and AssistanceProgram, 2004] and studies have shown that at even muchlower levels, particulate matter contributes to visibilityproblems and are likely responsible for respiratory andcardiopulmonary diseases like asthma, bronchitis, and heartdisease [Dockery et al., 1993; Pope et al., 2002; Agarwal etal., 2002; Chhabra et al., 2001; Yu et al., 2001]. Effective

strategies to mitigate such a problem rely on characterizingand identifying the sources of the PM that cause the highestexposure to humans [Smith, 1993]. Over the past severaldecades, a series of methods have been developed fordiagnosing the relative importance of the various emissionssources leading to fine particle air pollution problems inurban areas. Source-oriented methods of analysis exist thatrely on atmospheric transport models driven by detailedemissions inventories. However, this approach has beeninhibited by the lack of chemically speciated emissioninventories. A different approach to identifying sourceimpacts is to use receptor-based techniques which rely onobservations at sampling sites (receptors) and key tracerspecies. This method uses the differences in chemicalcomposition of particulate matter emitted from differentsources to identify the presence of particles from specificsources. The receptor-based method has been widely usedworldwide [e.g., Chow et al., 1992; Zeng and Hopke, 1989]and are particularly attractive for application in regions thathave not been studied in detail because they are able to yieldrapid insights into the causes of a local air pollution problembefore the completion of an accurate emissions inventory. Inthe Indian subcontinent, such a receptor-based sourceapportionment has been conducted for Bangladesh and inIndia using inorganic elemental analysis of PM2.5 [Begum et

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112, D15303, doi:10.1029/2007JD008386, 2007

1School of Earth and Atmospheric Sciences, Georgia Institute ofTechnology, Atlanta, Georgia, USA.

2Currently at Environmental Health Sciences, School of Public Health,University of California, Berkeley, California, USA.

3Environmental Chemistry and Technology Program, University ofWisconsin-Madison, Madison, Wisconsin, USA.

4Environmental Engineering Sciences, California Institute of Technol-ogy, Pasadena, California, USA.

5Department of Civil and Environmental Engineering, Georgia Instituteof Technology, Atlanta, Georgia, USA.

6Deceased 30 July 2001.

Copyright 2007 by the American Geophysical Union.0148-0227/07/2007JD008386

D15303 1 of 14

al., 2004; Kulkarni, 2006]; however, no receptor-basedsource apportionment work using organic compounds astracers has been conducted in this region. A significantfraction of the fine particulate mass in the urban atmosphereis organic and organic compounds present in fine particlesemitted from burning wood, combustion of automotive fuels,and combustion of coal are very different and can be exploitedto determine their respective contributions to the atmosphericconcentrations of fine particulate matter, as demonstrated inLos Angeles by Schauer et al. [1996] and in the southeastUnited States by Zheng et al. [2002]. Here, this receptor-

based organic tracer method has been used to quantify thesources that contribute to PM2.5 at four cities in India.

2. Experimental Setup

[3] Ambient sampling over a 1-year period was con-ducted in Delhi, Mumbai, and Kolkata (Figure 1), threemegacities located in India. To understand the seasonalpattern observed in the region, samples were taken duringMarch/April (spring), June/July (summer), October/November(autumn), and December/January (winter). For each of the fourseasons sampled, samples were taken every 6th day until atleast five to six samples were obtained for that season in thatsite. After analyzing 5 years of backwind trajectories fromNOAA [Draxler and Rolph, 2003; Rolph, 2003], Chandigarh,upwind of Delhi, was chosen as a fourth site representingbackground characteristics. It is a smaller citywith a populationof 809,000 (considered small in the Indian context), locatednorth of Delhi. Five samples during the summer season wereobtained at Chandigarh. There are very few data sets in theselarge cities that represent significant exposures tomany people,and thus even small data sets like this can be useful incharacterizing air pollution.[4] Sites at Delhi, Mumbai, and Kolkata were selected to

avoid undue influence of emissions coming from nearbycity traffic or industries, yet were located within the metro-politan area (see Table 1). In each location, a four-channel,PM2.5 filter sampler (see Figure 2) was placed either on arooftop or in the middle of open field to ensure that thesampler inlet was able to sample particles coming fromall directions. The Chandigarh site was located outside ofthe town. Sampling commenced on 4 March 2001 andcontinued until 16 January 2002. Samples were collectedat ambient temperatures and relative humidities for 24 hoursstarting at midnight local time every 6th day for each of themonths sampled, leading to 21 samples for Delhi, 25 forMumbai, 20 for Kolkata, and 5 for Chandigarh.[5] Fine particulate matter was collected on one quartz

fiber filter (Pallflex, 2500 QAO, 47 mm diameter), twoprewashed Nylon Filters (Gelman Sciences, Nylasorb,47mm diameter), and on two PTFE filters (Gelman Sciences,Teflo, 1.0 mm pore size). Ambient air was drawn at a rate ofapproximately 22.5 lpm through an acid-washed Pyrexglass inlet line to a Teflon-coated Air and Industrial Hygiene

Figure 1. Location of the four sampling sites: Delhi,Mumbai, Kolkata, and Chandigarh.

Table 1. Description of the Sampling Sites

City Site Address in India Location Type Site Description Source of Pollution

Mumbai NEERI Zonal Lab 89/B,Dr. Annie Basen Rd.Worli, Mumbai 400018

UrbanResidential

Sampler placed 3 mabove ground on a rooftop.A four-story building andslum area nearby.

City traffic typically seen inresidential and business areas,and cooking by slum dwellers.

Delhi National Physical LabDr. K. S. Krishnan MargNew Delhi 110012

UrbanResidential

Sampler placed 5 mabove ground on rooftop ofoffice building in NPL campus.Unobstructed space around.

City traffic typically seen inresidential and business areas,and cooking by slum dwellers.

Kolkata NEERI Zonal LabI-8, Sect-C, East KolkataP.O. Box Haltu Kolkata700078

UrbanResidential

Sampler on a 2 m platformlocated in open field.Ruby General Hospital anda diesel truck garage nearby.

City traffic typically seen inresidential and business areas,cooking by slum dwellers, andsome emission from combustion bydiesel trucks parked in nearby garage.

Chandigarh Postgraduate Institute ofMedical Education andResearch Sector_12Chandigarh 160012

Background Sampler on rooftop locatedon fourth floor. Buildinglocated in middle of campusaway from streets.

Road dust, gasoline vehicle trafficseen in planned town settings.

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

2 of 14

D15303

Laboratory (AIHL)-design cyclone separator [John andReischl, 1980], which removed large particles with a collec-tion efficiency curve having a 50% aerodynamic cutoffdiameter at 2.5 mm before the air passed through the filters.The nylon filter located downstream of the MgO-coateddiffusion denuder was used in conjunction with the nylonfilter downstream of the cyclone alone to measure gas-phasenitric acid, hydrochloric acid, and fine particle nitrate by thedenuder difference method. The air flow rate through eachfilter was measured before and after each 24-hour samplingperiod with a calibrated rotameter.[6] One of each pair of PTFE filter samples was analyzed

by ion chromatography (Dionex Corp, Model 2020i) for theanions NO3

�, SO42�, and Cl� [Mulik et al., 1976] and by an

indophenol colorimetric procedure for NH4+ [Bolleter et al.,

1961] using an Alpkem rapid flow analyzer (Model RFA-300). The second set of each of these sample sets wasanalyzed for trace elements using X-Ray Fluorescence(XRF) by the Desert Research Institute (DRI). Quartz fibersubstrates were analyzed for elemental and organic carboncontent using the thermal-optical carbon analysis method ofHuntzicker et al. [1982] as modified by Birch and Cary[1996]. In the thermal evolution and combustion method of

Birch and Cary [1996], elemental carbon is defined ascarbon that resists volatilization up to a temperature of900�C in an inert atmosphere. In this paper we will usethe term elemental carbon (EC) to define the carbondetected by this method. Variations in elemental carbonvalues between alternative methods can arise due to differ-ences in the way that alternative methods correct forcharring of the samples during analysis.[7] Extraction of particle-phase organic compounds col-

lected on quartz fiber filters was based on the methodsdescribed by Mazurek et al. [1987] and further refined bySchauer et al. [1996] and Zheng et al. [2002]. Samples werecombined by season and extracted in annealed glass jarswith Teflon-lined lids. Mumbai summertime samples didnot contain enough organic carbon (OC) for acceptable GC/MS analysis and thus were not analyzed. In the combinedsample, approximately 0.5 mg of OC is desired for accurateGCMS analysis, but summertime Mumbai total OC wasmuch less. In addition, filter blanks, as well as lab blanks,were analyzed. Filter blanks were prepared, stored, shippedin the same manner as the samples, and lab blanks wereused to identify possible contaminants from handling sam-ples in the laboratory. Both field and lab blanks were

Figure 2. Schematic diagram of the sampling unit. TU1 and TU2 are teflon filters, NU1 and ND1 areundenuded nylon filter and denuded nylon filter, respectively, and QU1 is a quartz fiber filter.

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

3 of 14

D15303

analyzed and levels from field blanks were subtracted fromthe sample, whereas the lab blanks were used to see if therewere any interferences. In total, 23 of the combined sampleswere analyzed, including field and laboratory quality assurance/quality control (QA/QC) blanks.[8] Samples were first spiked with an internal standard

mix containing 16 deuterated compounds. Deuterated inter-nal standards were dodecane-d26, hexadecane-d34,eicosane-d42, octacosane-d58, hexatriacontane-d74, benz-aldehyde-d6, decanoic acid-d19, heptadecanoic acid-d33,phthalic acid-3,4,5,6-d4, acenaphthene-d10, 4,40-dimethox-ybenzophenone-d8, dibenz(ah)anthracene-d14, chrysene-d12, aaa-20R-cholestane-d4, cholesterol-2,2,3,4,4,6-d6,and levoglucosan-13C6 (carbon-13 uniform-labeled com-pound). These standards provided internal quantificationreferences for the key particle phase organic compoundscovering their range of mass spectral fragmentations, polar-ity, and reactivitiy with derivatization reagents. About250 mL of the internal standard mix was spiked permilligram of OC. The amount of spiked internal standardmix was proportional to the amount of the OC present inthe sample. Samples were extracted using mild sonication(20 min) twice with hexane (Fisher Optima Grade),followed by three successive extractions using a 2:1mixture of benzene and isopropanal (benzene: HPLCGrade, E&M Scientific Omnisolv; isopropanal: FisherOptima Grade). Benzene was further distilled in thelaboratory to remove the small fraction of impurities andtested for purity by GC/MS prior to use. The extract wasfiltered to remove loose filter materials, and the volumewas reduced to about 5 mL using a rotary evaporator.

Finally, it was blown down to the volume of injectedinternal standard using ultrapure N2. The extract was splitinto two fractions. One fraction was derivatized withdiazomethane to convert organic acids to their methylester analogues for improved quantification.[9] Following Zheng et al. [2002], a Hewlett-Packard

GC/MSD (6890 GC and 5973MSD) equipped with a 30-mlength � 0.25-mm i.d. � 0.25-mm film thickness HP5 MScapillary column was used. Operating conditions wereisothermal hold at 65�C for 2 min, temperature ramp of10�C min�1 to 300�C, isothermal hold at 300�C for 22 min,GC/MS interface temperature 300�C. The flow of the carriergas, He, was 1 mL min�1. Injection volume was 1 mL foreach sample. Scan range was 50–500 amu, and the samplewas analyzed under electron ionization mode (70 eV).Estimated measurement uncertainties of the organic com-pounds were ±20% (1 sigma) [Schauer et al., 1999a].[10] Chemical Mass Balance (CMB 8.0) [Watson et al.,

1998] modeling was used to apportion PM2.5 particles tosources. The CMB model combines chemical and physicalcharacteristics of particles or gases measured at the sourcesand the receptors to quantify the source contributions to thereceptor [Miller et al., 1972]. In the present study, CMB wasconducted using organic compounds as molecular markers.An important aspect of molecular marker source apportion-ment is the selection of organic compounds that can beproperly used as tracer species in the model. Here 29organic compounds, along with Al, Si, and elementalcarbon (EC) were selected for use as tracer species in theCMB model (see Table 2). These species do not form, donot significantly react, nor have other selective removalprocesses (i.e., volatilization) in the atmosphere [Schauer

Table 2. Average Seasonal Concentrations of 29 Organic Compounds Used in the CMB Model, ng-m�3

Organic CompoundsDelhiSpring

DelhiSummer

DelhiAutumn

DelhiWinter

KolkataSpring

KolkataSummer

KolkataAutumn

KolkataWinter

MumbaiSpring

MumbaiAutumn

MumbaiWinter

ChandigarhSummer

n-Pentacosane 32.12 4.47 37.55 102.43 18.78 10.92 23.69 117.92 5.84 18.70 21.59 23.31n-Hexacosane 32.14 6.19 36.17 85.92 14.54 19.89 14.11 90.70 7.48 20.48 20.68 14.30n-Heptacosane 49.19 9.62 54.32 105.10 22.32 15.81 24.71 103.57 12.04 22.71 21.89 24.64n-Octacosane 37.23 7.01 43.45 76.67 13.68 17.70 19.13 73.75 8.50 24.31 16.43 11.64n-Nonacosane 86.17 16.79 77.38 150.32 27.37 18.84 45.29 86.32 24.13 31.64 23.58 13.03n-Triacontane 16.60 3.63 23.50 33.44 12.19 16.69 29.11 31.71 3.73 17.88 7.72 3.95n-Hentriacontane 39.36 7.04 43.02 71.70 12.56 8.59 10.31 61.99 8.08 32.95 12.92 6.92n-Dotricontane 31.92 8.38 53.27 94.62 16.09 27.10 24.06 81.76 14.07 9.97 20.02 78.34n-Tritriacontane 47.12 10.89 74.32 121.76 12.26 14.23 29.36 105.14 11.61 13.94 23.22 5.0820S&R-abb-Cholestanes

0.81 0.24 1.08 2.99 1.51 0.50 2.43 2.98 0.29 0.49 0.77 1.97

20R-aaa-Cholestane 0.72 0.19 0.66 2.09 0.92 0.70 2.31 5.34 0.18 0.28 0.61 0.5020S&R-abb-Ergostanes 1.95 0.32 1.04 4.30 1.22 0.65 2.10 9.19 0.41 0.98 2.41 1.4020S&R-abb-Sitostanes 3.23 0.73 3.80 8.05 1.51 1.22 4.30 9.39 1.04 1.80 2.84 3.6622, 29, 30-Trisnorneohopane

0.92 0.53 2.19 5.35 1.72 0.45 2.29 8.97 0.30 0.03 1.10 3.28

17a, 21b-29-Norhopane 4.75 1.56 9.89 17.37 5.43 3.28 5.30 21.97 1.20 2.59 5.06 8.9317a, 21b-Hopane 3.83 1.65 7.37 15.42 6.74 4.01 8.52 19.98 0.99 1.42 3.66 11.05Isopimaric Acid 2.93 1.12 7.36 28.67 4.03 2.06 6.43 21.06 3.58 3.23 5.59 5.74Hexadecanamide 17.40 4.14 18.96 44.58 7.62 2.93 6.90 36.59 4.32 2.76 6.97 5.02Octadecanamide 6.87 1.35 7.55 15.32 2.31 1.13 3.98 9.36 1.30 1.18 2.15 1.65Benzo[b]Fluoranthene 7.11 1.26 9.97 30.62 4.48 1.67 6.07 53.59 1.25 3.95 6.71 0.68Benzo[k]Fluoranthene 6.94 1.13 8.10 28.39 4.36 1.31 5.10 40.86 1.25 3.48 5.59 0.53Benzo[e]Pyrene 6.39 0.82 7.54 25.77 4.19 0.88 5.01 39.38 0.45 2.13 2.69 0.40Indeno[1,2,3-cd]Fluoranthene

3.51 0.92 5.07 16.58 2.44 1.08 3.36 23.41 0.81 2.26 3.26 0.43

Indeno[1,2,3-cd]Pyrene 3.90 1.06 6.57 18.49 2.85 1.61 3.85 26.13 0.95 2.32 3.58 0.49Picene 1.26 0.28 1.54 5.11 0.66 0.27 0.82 7.12 0.14 0.48 0.93 0.16Coronene 6.51 1.64 11.16 21.60 3.58 1.70 3.95 35.95 0.76 2.43 5.13 0.65Stigmasterol 77.26 29.20 164.40 295.86 49.81 21.79 36.85 301.28 10.33 142.54 34.61 60.05Levoglucosan 1026.60 210.46 1773.64 5258.30 336.45 75.12 474.38 5491.95 74.52 392.43 907.96 140.30

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

4 of 14

D15303

et al., 1996]. Eight source profiles (see Table S1 andFigures S1a and S1b in auxiliary material) were selected foruse in this study.1 These source profiles were developedusing the same laboratory procedures described previouslyfor atmospheric samples. Whenever possible, source pro-files from the Indian subcontinent were selected. Fivesource profiles were from Bangladesh (combustion ofcoconut leaves, rice straw, cow dung, jackfruit wood, andbiomass briquette) [Sheesley et al., 2003]. Given the sim-ilarity of these profiles, leading to colinearity problems, acombined biomass burning source profile was developedusing equal amounts from four of the five (coconut leaves,rice straw, jackfruit wood, and biomass briquette), while thecow dung source profile, which was slightly different fromthe other four, was kept as a separate source profile. Sourceprofiles for diesel engine exhaust, gasoline engine exhaust,fuel oil, and road dust were obtained from previous studiesin North America [Hildemann et al., 1991; Rogge et al.,1993a, 1993b, 1993c, 1997; Schauer et al., 1999b, 2001,2002b]. The coal source profile was obtained from theanalysis of fine particulate matter emitted from the burningof Datong coal in China [Zheng et al., 2005]. While not all

of the source profiles have been directly measured in India,the chosen tracer species are more specific to the sourcesthan to the locations.

3. Results and Discussion

3.1. Organic Speciation Results

[11] Samples from the four cities show a seasonallyvarying distribution of organic marker compounds (seeTable 2, Figures 3a and 3b and also Figures S2a–2c inthe auxiliary material), indicating a similarly varying set ofsources. Hopanes and steranes are organic markers that arepresent in heavy petroleum distillates such as lubricating oil[Simoneit, 1985, 1999]. In the southern California atmo-sphere, these compounds have been shown to be predom-inately from the gasoline exhaust and diesel-powered motorvehicles, resulting from the presence of lubricating oil[Rogge et al, 1993a, 1993b, 1996; Schauer et al., 1996;Schauer et al., 2002a]. Diesel vehicles are important sourcesof both elemental carbon and hopanes and steranes, whilegasoline-powered vehicles are smaller contributors to ele-mental carbon concentrations. On the other hand, levoglu-cosan is a major component of wood smoke aerosol and hasbeen shown to be a good tracer for biomass burning

Figure 3a. Seasonal variations of elemental carbon and organic carbon for Delhi, Kolkata, Mumbai,and Chandigarh.

1Auxiliary material data sets are available at ftp://ftp.agu.org/apend/jd/2007jd008386. Other auxiliary material files are in the HTML.

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

5 of 14

D15303

[Schauer et al., 1996; Simoneit et al., 1999] but does notcontain hopanes and steranes. Silicon and aluminum aremarkers for road dust. Picene is used as a tracer for coalcombustion [Oros and Simoneit, 2000], and stigmasterolcombined with the amides (Hexadecanamide and Octade-canamide) are suggested as tracers for cow dung combus-tion [Sheesley et al., 2003].[12] Summertime levoglucosan concentrations for Delhi,

Kolkata, and Chandigarh were 210 ± 40 ng/m3, 75 ± 15 ng/m3,and 140 ± 30 ng/m3, respectively, whereas wintertimelevoglucosan concentrations for the Delhi, Kolkata, andMumbai are 5300 ± 1100 ng/m3, 5500 ± 1100 ng/m3, and910 ± 180 ng/m3, respectively. As explained earlier, sum-mertime Mumbai measurements did not meet the criteriafor analysis by GC/MS; on the other hand, wintertimeChandigarh measurements were not taken in this study.Ratios for levoglucosan to the sum of hopanes and steranesfollow a seasonal trend: ratios are at least three to 10 timeshigher during the colder months compared to the warmermonths for all the cities in this study. This seasonal trend inlevoglucosan may suggest increased biomass use for home

heating during winter or rice straw burning in the field in thefall. Stigmasterol as well as Hexadecanamide and Octade-canamide, used to identify cow dung smoke, have beendetected in all cities. Low-income households use cow dungpatties along with tree leaves and branches to cook and heat.No food cooking source profile applicable for this regionwas available and it is possible that some of the stigmasterolmay be attributed from food cooking operations.[13] Picene concentrations in the summer in Delhi, Kol-

kata, and Chandigarh were 0.30 ± 0.06 ng/m3, 0.30 ±0.06 ng/m3, and 0.20 ± 0.04 ng/m3, respectively, andwintertime concentrations in Delhi, Kolkata, and Mumbaiwere 5.1 ± 1.0 ng/m3, 7.1 ± 1.4 ng/m3, and 0.9 ± 0.2 ng/m3,respectively. Concentrations of picene increased during thewinter, likely because of an increase in the use of coal forresidential heating. Three thermal power plant stations arepresent in Delhi: Indraprastha (284 MW capacity, burning1,150,000 MT/yr of pulverized coal with 39.4% ash and0.36% sulfur content in the coal), Rajghat (135 MWcapacity, burning 876,000 MT/yr of pulverized coal with

Figure 3b. Seasonal variations of levoglucosan, picene, hopanes and steranes for Delhi, Kolkata,Mumbai, and Chandigarh. The sum of hopanes and steranes is designated by (H + S).

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

6 of 14

D15303

Figure 4. Comparison of the calculated and measured ambient concentrations of the mass balancespecies used in the CMB model in (a) Delhi and (b) Kolkata during autumn 2001.

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

7 of 14

D15303

a 35–42% ash and 0.50% sulfur content in the coal), andBadarpur (720 MW capacity, burning 3,940,000 MT/yr ofpulverized coal with a 28–32% ash and 0.35% sulfurcontent) [Central Board of Irrigation and Power, 1997].Although electrostatic precipitators (ESPs) are present ineach of these power generating facilities, poor maintenancehas often been blamed for high emissions of fly ash[Aggarwal et al., 1999]. Cholesterol concentrations werebelow the instrument detection limit in almost all thesamples, suggesting that the amount of meat cooking isminimal, consistent with a large fraction of Indians beingvegetarian.[14] We see reasonably good agreement when comparing

our PM2.5 Delhi average wintertime organic carbon speci-ation results (Table 2) with the PM10 wintertime organicspeciation results from the pilot work conducted at the IndiaHabitat Center (IHC) in New Delhi by Sharma et al. [2003].The concentration found at our site is 3–52% higher for n-alkanes, relatively close for the PAHs, and three timeshigher for levoglucosan.

3.2. CMB Results

[15] The source contribution to organic carbon can becomputed by CMB modeling method as described bySchauer et al. [2002a] and Zheng et al. [2002]. The CMBresults presented in this study are statistically significantwith r2 ranging from 0.82 to 0.95. As seen in Figures 4aand 4b, there is good agreement between the calculatedconcentrations for each of the CMB fitting species

(29 organic species along with EC, Al, and Si) and themeasured concentrations for the same species. Figure 5shows the five identified sources (diesel, gasoline, coal,biomass, and road dust) that contribute to the organiccarbon in PM2.5. The ‘‘other organics’’ category listed inthe figure represents the residual difference between themeasured total organic carbon concentrations and the sumof the apportioned contributions of the five identifiedsources as quantified from the receptor model.[16] Since the ratio of the emissions of fine organic

carbon to fine particle mass is known for each of thesources in the model, source contributions to fine particlemass concentrations can be computed. Figure 6 and Table 3show the fine particle mass contributions from the primarysources plus the unapportioned material as well as second-ary sulfate, nitrate, and ammonium ion concentrations.From Figure 6 we can see that there is no single dominantsource of PM2.5. Gasoline combustion is primarily frommobile sources, but the diesel contribution is likely fromboth stationary and mobile sources. It is not possible toattribute secondary sulfates, nitrates, and ammonium tospecific primary sources using CMB, although sulfatescan likely be linked to the sulfur in fossil fuels. Secondarysulfates, nitrates, and particle-phase organics formationcomprised approximately one-tenth to one-fifth of PM2.5.Broadly, mobile sources and biomass combustion appear tocontribute substantially and in several cases approximatelyin equal proportions, while road dust dominates at times.

Figure 5. Seasonal source contribution to organic carbon in PM2.5 in Delhi, Kolkata, Mumbai, andChandigarh during 2001.

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

8 of 14

D15303

Figure 6. Source contribution to the ambient fine particles in Delhi, Kolkata, Mumbai, and Chandigarh.The sizes of the pie charts are proportional to the average seasonal PM2.5 concentration.

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

9 of 14

D15303

Predictably, the contributions of biomass and coal, presum-ably used for heating, are high in winter in Delhi andKolkata.[17] During INDOEX high EC to OC ratios were ob-

served in the small island of Kaashidhoo where the inten-sive field experiment occurred [Chowdhury et al., 2001] andover the Indian Ocean from ship cruises [Ramanathan et al.,2001] suggesting a strong heating of the atmosphere fromaerosol. The high EC concentrations have been relatedthrough trajectory analysis to the source regions in theIndian subcontinent [Chowdhury et al., 2001; Ramanathanet al., 2001]. The current study conducted upwind of theINDOEX region confirms these high EC to OC ratios.Moreover, the combustion of solid biomass, such as wood,agricultural waste, and dried animal manure in cooking stoves,is the largest source of black carbon emissions in India[Venkataraman et al., 2005]. EC/OC ratios of 0.11 to 3.53in Indian biomass aerosol were reported by Venkataramanet al. [2005] which are much higher than previous reportedratios. Kolkata, located in the eastern part of India wheremost of the rural people use biomass and cow dung forcooking and heating, shows high EC to OC ratios.[18] Unattributed mass (‘‘Other Mass’’ category) was also

substantial in some of the samples. This fraction is due, inpart, to unidentified organics and water. A fraction of theunidentified OC can be due to secondary formation. PM2.5

diesel exceeds gasoline in almost all cases, which is notsurprising, given the relatively higher emission rates fordiesel compared to gasoline and the higher consumptionof diesel compared to gasoline in India. The source profilefor diesel is from medium-duty diesel trucks, and it is hardto distinguish between mobile and stationary diesel com-bustion. The use of diesel in small power generators is notinsignificant in the Indian cities studied because of frequentpower outage, so it is likely that not all diesel-derived PM2.5

is from mobile sources. Gasoline, in contrast, is used almostexclusively in vehicles, and can be attributed to mobilesources with little error. By summing the contributions fromdiesel, gasoline, and coal, it can be seen that in most casesfossil fuel combustion exceeds biomass combustion. Thereconstructed mass from identified sources from the CMBresults for summer in Kolkata is 107% of the measured fineparticle mass. Uncertainties in the reconstructed mass, asseen in Table 3, and uncertainties in the measured fineparticle mass can explain this discrepancy. The precision ofthe mass measurement is on the order of 1 to 2 mg/m3. Also,higher levels can be due to larger ratios of total PM2.5 to OCfrom source emissions than are actually present, and prop-agation of the various uncertainties involved in this process[Zheng et al., 2002], or the use of an organic matter toorganic carbon ratio (a ratio of 1.7 was used in here toconvert OC to organic matter) higher than actually present.Total mass concentrations from identified sources shouldequal approximately 100%, although values ranging from80 to 120% are acceptable [Watson et al., 1990].

3.3. Fuel-Based Particulate Matter Emissions

[19] An initial confirmation of the CMB-based sourceapportionment was conducted by comparing the results herefor fuel oil, diesel, and gasoline emissions to the resultsobtained from National Environmental Engineering Re-search Institute’s (NEERI) yearly fuel usage and fuel-basedT

able

3.CalculatedConcentrationsandUncertainties,mg-m

�3,ofSources

andPerform

ance

ofCMBAnalysisbySeasonandSite

Delhi

Kolkata

Mumbai

Chandigarh

Spring

Summer

Autumn

Winter

Spring

Summer

Autumn

Winter

Spring

Autumn

Winter

Summer

Diesel

20.00±2.14

10.66±1.28

23.39±2.80

23.03±17.17

10.82±1.45

5.10±1.58

12.27±8.26

50.16±5.63

8.32±0.99

11.78±1.41

18.95±1.94

3.68±0.43

Gasoline

5.48±0.60

0.90±0.19

6.19±0.73

20.68±2.35

6.09±0.71

4.88±0.35

6.54±1.31

32.15±3.49

0.99±0.17

0.96±0.25

5.71±0.58

4.61±0.61

Road

Dust

25.38±1.19

20.77±0.95

36.68±1.35

24.43±0.95

15.12±0.72

5.57±0.40

1.99±0.72

21.80±1.27

13.69±0.64

14.72±0.95

22.60±0.72

9.23±0.48

Coal

7.73±0.93

0.83±0.20

9.58±1.34

32.57±4.08

3.42±0.59

1.17±0.22

4.30±0.76

44.01±5.77

0.78±0.15

0.86±0.56

6.87±0.73

0.00±0.00

BiomassBurning

18.34±3.77

3.22±0.91

25.35±4.95

45.87±9.64

7.09±1.60

3.73±0.91

8.11±1.83

40.78±9.39

2.63±0.67

12.74±2.50

11.05±1.85

2.32±0.21

SecondarySulfate

8.98±0.14

4.80±0.12

9.22±0.22

17.84±0.65

8.36±0.14

2.86±0.12

3.76±0.10

12.00±0.31

5.33±0.12

7.54±0.16

10.37±0.26

4.58±0.13

SecondaryNitrate

2.29±0.16

1.38±0.14

3.35±0.26

16.51±0.76

0.95±0.12

0.78±0.14

0.45±0.12

8.75±0.36

0.59±0.11

1.76±0.14

2.69±0.16

0.66±0.15

SecondaryAmmonium

6.39±1.89

1.41±0.05

3.27±0.10

11.06±0.31

1.72±0.05

0.34±0.05

1.13±0.05

9.06±0.14

0.81±0.04

1.84±0.06

3.46±0.06

1.71±0.06

Other

Mass

19.60±5.72

5.53±1.98

42.06±6.09

38.91±20.36

1.12±2.95

2.07±1.97

6.15±8.64

85.78±12.98

2.77±2.71

11.70±3.91

7.20±2.99

2.43±1.17

TotalCalculatedMass

94.60±5.01

43.97±1.87

117.04±6.05

191.99±20.30

53.58±2.46

24.43±1.92

38.55±8.62

218.72±12.93

33.13±1.38

52.20±3.09

81.70±2.94

26.79±0.93

TotalMeasuredMass

114.2

±2.76

49.5

±0.64

159.1

±0.63

230.9

±1.6

54.7

±1.63

26.5

±0.45

44.7

±0.56

304.5

±1.13

35.9

±2.33

63.9

±2.4

88.9

±0.54

29.22±0.7

r20.93

0.94

0.93

0.92

0.95

0.92

0.92

0.92

0.92

0.90

0.92

0.82

c2

3.03

2.7

2.91

3.45

1.99

3.19

2.92

3.39

3.38

4.45

3.64

6.03

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

10 of 14

D15303

emissions factors in Mumbai for the year 2001. Emissionfactors were obtained from Reddy and Venkataraman[2002]. The emission factor for gasoline vehicles is aweighted average between the emission factor for non-catalytic vehicles using unleaded gasoline [Reddy andVenkataraman, 2002] and for motorcycles [EnvironmentalProtection Agency, 2002]. Diesel contribution to fine parti-cle emissions in Mumbai is an order of magnitude higherthan gasoline and fuel oil (Table 4). This is comparable toour results obtained from CMB. In addition, gasoline anddiesel emissions from vehicular activity are directly emittedat ground level and thus have a proportionally greater effecton urban air quality. Reddy and Venkataraman [2002] findthat utility coal burning has the largest emissions nationallyin India. Such emissions are not concentrated as much incities as are motor vehicle emissions. Further, those emis-sions, along with the emissions from burning fuel oil, areoften injected into the atmosphere well above the mixedlayer. Thus one expects fuel oil and coal sources to have asmaller impact, relative to their total emissions rate, onurban, ground level particulate measurements, as foundhere.[20] It is also of interest to conduct an order of magnitude

approximation of the expected levels of PM from thevarious sources. Using an approximate size of greaterMumbai of 50 � 50 km2, an average mixing height of200 m, and an average wind speed of 2 m/s, and assuminginstant mixing in to the whole airshed, the above emissionsestimates would suggest PM levels from the three sourceswould be about 5 mg/m3 for diesels, 0.3 mg/m3 for gasolinefueled vehicles, and 0.2 mg/m3 for furnace oil. While all ofthese values appear low (due to the approximations in thecalculation and possible underestimates in the emissionsfactors for the various sources), it does suggest that it is notsurprising that our source apportionment does not findsignificant levels of PM coming from furnace oil, kerosene,and other liquid and gaseous fuel used in industrial, externalcombustion boilers. Again, gasoline fueled vehicles emitnear the ground, similar to where the monitors are samplingair, and will have a greater impact, particularly at nightwhen the mixing depths are much lower.

4. Uncertainties

[21] Receptor modeling of the type performed here isopen to uncertainties, though the use of organic molecularmarkers provides significant extra information. Further, theagreement with other approaches of estimating sources ofPM2.5 conducted above is encouraging, as is the agreementwith other studies [Reddy and Venkataraman, 2002]. Whenpossible, we use source profiles specific to the region,

recognizing that a source in one location can have adifferent profile than ostensibly the same source in anotherlocation, for example, diesel trucks. The source profile used,in that case, was developed in the United States. The profilefor coal was taken from a Chinese coal. While a compre-hensive formal error analysis is beyond the scope of thiswork, Table 3 does provide calculated uncertainties basedon data provided in the CMB analysis. S. Lee and A.Russell (manuscript in preparation, 2007) conducted MonteCarlo analysis with Latin hypercube sampling to betterunderstand the uncertainties of the source contributionsapportioned by CMB using inorganic species. The objectivewas to evaluate the source contribution uncertainties and toidentify uncertainty contributors due to uncertainties inambient measurement and source profiles. They found thatthe uncertainties in the source profiles contribute more tothe source contribution uncertainties as compared to theuncertainties in ambient measurement data. Zheng et al.[2006] conducted sensitivity tests for biomass and road dustby replacing the key organic tracers’ concentrations anduncertainties with the concentrations and uncertaintiesfound in different profiles for the same source whilekeeping all other species the same. They show that thebiomass relative contribution changed 63–130% and roaddust relative contribution changed 8–19%. In the presentpaper, sensitivity analysis was conducted on wintertimeDelhi CMB results by changing the uncertainties of thekey tracers for each of the source profiles used in the CMBwork. This approach provides an assessment of the sensi-tivity of the results to the uncertainty choice used, but likelygives less variation than if one uses a range of sourceprofiles [e.g., Zheng et al., 2006]. Uncertainties of the keytracers for one profile were first increased by 25%, then by50%, and finally by 100% while keeping the uncertaintiesof all remaining species and all other source profilesconstant (see Table 5). This was repeated for all sourceprofiles and the CMB work was conducted and comparedwith the base scenario. Increasing the uncertainties usuallyhad a small effect on the calculated uncertainties and on thefinal source apportionment results (average CoV of about4%). The diesel and gasoline contribution appeared to bethe only two sources being impacted significantly whenchanging the uncertainties of the hopanes and steranes. Forother sources, the uncertainties in both the apportioned massand the CMB-derived uncertainties are small, suggestingthat the uncertainties and the results are robust.

5. Conclusions

[22] Chemically detailed particulate matter characteriza-tion, including organic speciation, and detailed source

Table 4. First-Order Approximation of Particulate Emission From Mumbai Using a Fuel-Based Approacha

YearlyConsumption

EmissionFactor Density

Fuel-BasedPM Emission

CMB PMEmission

Reddy andVenkataraman

106 l g/kg kg/l Kg/Day % of Fine Entire IndiaFurnace Oil 424 0.65 0.95 620 Negligible NegligibleDiesel (automotive+ industrial)

1140 4.2 0.85 11,000 21% 10%

Gasoline 565 0.6 0.75 700 3.6% NegligibleaThe estimate for gasoline is sensitive to the assumed fraction of fuel use by motorcycles and the emissions factor for that

source.

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

11 of 14

D15303

Table5.SensitivityAnalysisonWintertim

eDelhiSourceApportionment(TotalMeasuredMass230.9

mgm

�3)byMultiplyingtheUncertainties

ofKey

Tracers

by1.25(25%

Increase),1.5

(50%

Increase),and2(100%

Increase)a

Diesel

Gasoline

Road

Dust

Coal

BiomassBurning

Sec.Sulfate

Sec.Nitrate

Sec.Ammonium

Others

TotalCalculated

MassConc.

BaseCase

23.03±17.17

20.68±2.35

24.43±0.95

32.57±4.08

45.87±9.64

17.84±0.65

16.51±0.76

11.06±0.31

38.91±20.36

191.99±20.30

Diesel�

1.25

23.45±17.17

20.09±2.35

24.51±0.95

32.68±4.08

45.25±9.64

17.15±0.12

16.52±0.14

11.06±0.05

40.19±20.33

190.71±20.27

Diesel�

1.5

23.51±17.20

20.08±2.34

24.51±1.67

32.68±4.08

45.25±9.64

17.15±0.22

16.52±0.16

11.06±0.10

40.14±20.41

190.76±20.34

Diesel�

223.64±17.27

20.08±2.34

24.51±1.67

32.68±4.08

45.25±9.64

17.15±0.65

16.52±0.26

11.06±0.31

40.01±20.49

190.89±20.42

Gasoline�

1.25

17.78±16.48

21.52±2.70

24.12±1.59

32.41±4.08

45.27±9.66

17.21±0.14

16.53±0.12

11.11±0.05

44.95±19.85

185.95±19.78

Gasoline�

1.5

14.91±16.18

22.70±3.10

23.88±1.67

32.21±4.08

45.25±9.66

17.24±0.12

16.54±0.14

11.14±0.05

47.03±19.67

183.87±19.60

Gasoline�

211.69±15.99

24.46±3.94

23.56±1.67

31.92±4.08

45.23±9.66

17.27±0.10

16.55±0.12

11.18±0.05

49.04±19.66

181.86±19.56

Road

Dust�

1.25

25.39±17.96

19.96±2.35

24.59±1.75

32.70±4.08

45.23±9.64

17.13±0.31

16.52±0.36

11.05±0.14

38.33±21.06

192.57±21.00

Road

Dust�

1.5

25.98±5.16

19.92±2.17

24.59±0.72

32.73±4.08

45.19±9.60

17.13±0.12

16.52±0.11

11.02±0.04

37.82±11.97

193.08±11.86

Road

Dust�

225.98±5.16

19.92±2.17

24.59±0.80

32.78±4.08

45.17±9.58

17.13±0.16

16.52±0.14

11.02±0.06

37.79±11.96

193.11±11.85

Coal

�1.25

23.75±16.97

20.26±2.35

24.44±1.67

34.14±4.38

45.46±9.73

17.17±0.26

16.53±0.16

11.02±0.06

38.13±20.32

192.77±20.26

Coal

�1.5

25.72±17.47

20.11±2.35

24.51±1.67

31.01±4.01

46.77±10.08

17.13±0.13

16.52±0.15

11.06±0.06

38.07±20.82

192.83±20.76

Coal

�2

23.45±17.17

20.09±2.35

24.51±0.95

32.68±4.08

45.25±9.64

17.15±0.13

16.52±0.15

11.06±0.06

40.19±20.33

190.71±20.27

Cowdung�

1.25

22.99±17.11

20.09±2.35

24.44±1.67

32.78±4.08

45.08±9.75

17.16±0.13

16.52±0.15

11.06±0.06

40.78±20.38

190.12±20.32

Cowdung�

1.5

22.63±17.07

20.09±2.35

24.44±1.67

32.85±4.08

44.94±9.85

17.16±0.13

16.53±0.15

11.06±0.06

41.20±20.40

189.70±20.33

Cowdung�

222.10±16.97

20.09±2.35

24.36±1.67

32.95±4.08

44.79±9.94

17.17±0.13

16.53±0.15

11.06±0.06

41.04±20.36

189.86±20.30

Biomass�

1.25

23.45±17.17

20.09±2.35

24.51±1.67

32.68±4.08

45.25±9.64

17.15±0.13

16.52±0.15

11.06±0.06

40.19±20.38

190.71±20.32

Biomass�

1.5

23.45±17.17

20.09±2.35

24.51±1.67

32.68±4.08

45.25±9.64

17.15±0.13

16.52±0.15

11.06±0.06

40.19±20.38

190.71±20.32

Biomass�

223.45±17.17

20.09±2.35

24.51±1.67

32.68±4.08

45.25±9.64

17.15±0.13

16.52±0.15

11.06±0.06

40.19±20.38

190.71±20.32

AvgContribution

22.44±3.72

20.59±1.16

24.40±0.27

32.62±0.57

45.32±0.41

17.20±0.16

16.52±0.01

11.07±0.04

40.75±3.07

190.15±3.07

AvgUncertainty

15.79±3.77

2.47±0.41

1.46±0.36

4.09±0.07

9.70±0.13

0.20±0.17

0.19±0.15

0.09±0.08

19.45±2.66

19.38±2.67

COV

Contrib.

0.17

0.06

0.01

0.02

0.01

0.01

0.00

0.00

0.08

0.02

aHerer2

was

0.92–0.93in

allcasesandc2was

3.4

±0.1.

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

12 of 14

D15303

apportionment for Delhi, Mumbai, Kolkata, and Chandigarhwere conducted using receptor-based chemical mass bal-ance modeling. Five major sources of primary PM2.5 werequantified: diesel exhaust, gasoline exhaust, road dust, coalcombustion, and biomass combustion. Important trends inthe seasonal and spatial patterns of the impact of these fivesources were observed. Primary emissions from fossil fuelcombustion (coal, diesel, and gasoline) contributed 25–33% of the PM2.5 in Delhi, 21–36% in Mumbai, 37–57%in Kolkata, and 28% in Chandigarh. These figures can becompared to the biomass combustion contributing 7–20%in Delhi, 7–20% in Mumbai, 13–18% in Kolkata, and 8%in Chandigarh. Road dust was also significant. These resultsare generally consistent with calculations of PM2.5 emis-sions, and an order of magnitude calculation of sourcecontributions. Analyses suggest that the split between lightand heavy duty vehicles is somewhat uncertain, but the totalfossil fuel impacts and those of other sources are morerobust.

[23] Acknowledgments. Funding for this work was provided by theWorld Bank, the U. S. EPA (agreements RD83107601 and RD83096001),Georgia Power, and the College of Sciences at the Georgia Institute ofTechnology. Sampling work was conducted with the assistance and coop-eration of Chandra Venkataraman at the Indian Institute of Technology-Bombay, A. P. Mitra and C. Sharma at the National Physical Laboratory inDelhi, and Rakesh Kumar and A. Biswas, respectively, at the NationalEnvironmental Engineering Research Institute (NEERI) in Mumbai and inKolkata. These latter institutions provided manpower for gathering samplesthroughout the year. In addition, the authors gratefully acknowledge theNOAA Air Resources Laboratory (ARL) for the provision of the HYSPLITtransport and dispersion model and/or READY Web site (http://www.arl.noaa.gov/ready.html) used in this publication.

ReferencesAgarwal, K. S., M. Z. Mughal, P. Upadhyay, J. L. Berry, E. B. Mawer, andJ. M. Puliyel (2002), The impact of atmospheric pollution on vitamin Dstatus of infants and toddlers in Delhi, India, Arch. Diseases Child, 87,111–113.

Aggarwal, A. L., N. Sunita, and S. Sen (Eds.) (1999), The Citizens’ FifthReport, Cent. for Sci. and Environ., New Delhi.

Begum, B. A., K. Eugene, S. K. Biswas, and P. K. Hopke (2004), Inves-tigation of sources of atmospheric aerosol at urban and semi-urban areasin Bangladesh, Atmos. Environ., 38, 1352–2310.

Birch, M. E., and R. A. Cary (1996), Elemental carbon-based method formonitoring occupational exposures to particulate diesel exhaust, AerosolSci. Technol., 25, 221–241.

Bolleter, W. T., C. T. Bushman, and P. W. Tidwell (1961), Spectrophoto-metric determinations of ammonium as indophenol, Anal. Chem., 33,592–594.

Central Board of Irrigation and Power (1997), Compendium of ThermalPower Stations, Publ. no. 23, vols. I and II, Malcha Marg, Chanakyapuri,New Delhi.

Chhabra, S. K., P. Chhabra, S. Rajpal, and R. K. Gupta (2001), Ambient airpollution and chronic respiratory morbidity in Delhi, Arch. Environ.Health, 56, 58–64.

Chow, J. C., J. G. Watson, D. H. Lowenthal, P. A. Solomon, K. L. Magliano,S. D. Ziman, and L. W. Richards (1992), PM10 source apportionment inCalifornia’s San Joaquin Valley, Atmos. Environ., Part A, 26, 3335–3354.

Chowdhury, Z., L. S. Hughes, L. G. Salmon, and G. R. Cass (2001), Atmo-spheric particle size and composition measurements to support light ex-tinction calculations over the Indian Ocean, J. Geophys. Res., 106,28,597–28,605.

Dockery, D. W., C. A. Pope, X. P. Xu, J. D. Spengler, J. H. Ware, M. E.Fay, B. G. Ferris, and F. E. Speizer (1993), An association between air-pollution and mortality in 6 United-States cities, N. Engl. J. Medicine,329, 1753–1759.

Draxler, R. R., G. D. Rolph (2003), HYSPLIT (HYbrid Single-ParticleLagrangian Integrated Trajectory) Model, NOAA Tech. Memo ERLARL-224, Air Resour. Lab., Natl. Oceanic. Atmos. Admin., Silver Spring,Md. (Available at http://www.arl.noaa.gov/ready/hysplit4.html)

Environmental Protection Agency (2002), Exhaust emission factors fornon-road engine modeling-spark ignition, EPA420-P-02-015, Nov.2002, NR-010c, Environ. Prot. Agency, Washington, D. C. (Availableat http://www.epa.gov/otaq/models/nonrdmdl/p02015.pdf)

John, W., and G. Reischl (1980), A cyclone for size-selective sampling ofambient air, J. Air Pollut. Control Assoc., 30, 872–876.

Hildemann, L. M., G. R. Markowski, and G. R. Cass (1991), Chemicalcomposition of emissions from urban sources of fine organic aerosol,Environ. Sci. Technol., 25, 744–759.

Huntzicker, J. J., R. L. Johnson, J. J. Shah, and R. A. Cary (1982), Parti-culate Carbon, Atmospheric Life Cycle, edited by G. T. Wolff and R. L.Klimisch, pp. 79–88, Springer, New York.

Kulkarni, M. M. (2006), Source apportionment of human exposure to par-ticulates in Mumbai, India, Aerosol Air Qual. Res., 6, 281–294.

Mazurek, M. A., B. R. T. Simoneit, G. R. Cass, and H. A. Gray (1987),Quantitative high-resolution gas chromatography and high-resolution gaschromatography/mass spectrometry analysis of carbonaceous fine aerosolparticles, Int. J. Environ. Anal. Chem., 29, 119–139.

Miller, M. S., S. K. Friedlander, and G. M. Hidy (1972), A chemicalelement balance for the Pasadena aerosol, J. Colloid Interface Sci., 39,165–176.

Mulik, J., R. Puckett, D. Williams, and E. Sawicki (1976), Ion chromato-graphic analysis of sulfate and nitrate in ambient aerosols, Anal. Lett., 9,653–663.

Oros, D. R., and B. R. T. Simoneit (2000), Identification and emission ratesof molecular tracers in coal smoke particulate matter, Fuel, 79, 536–715.

Pope, C. A., III, R. T. Burnett, and M. J. Thun (2002), Lung cancer,cardiopulmonary mortality, and long-term exposure to fine particulateair pollution, J. Am. Medical Assoc., 287, 1132–1141.

Ramanathan, V., et al. (2001), Indian Ocean Experiment: An integratedanalysis of the climate forcing and effects of the great Indo-Asian haze,J. Geophys. Res., 106, 28,371–28,398.

Reddy, M. S., and C. Venkataraman (2002), Inventory of aerosol and sul-phur dioxide emissions from India: I –Fossil fuel combustion, Atmos.Environ., 36, 677–697.

Rogge, W. F., M. A. Mazurek, L. M. Hildemann, G. R. Cass, and B. R. T.Simoneit (1993a), Quantification of urban organic aerosols at a molecularlevel—Identification, abundance and seasonal variation, Atmos. Environ.,Part A, 27, 1309–1330.

Rogge, W. F., L. M. Hildemann, M. A. Mazurek, G. R. Cass, and B. R. T.Simoneit (1993b), Sources of fine organic aerosol. 2. Non-catalyst andcatalyst-equipped automobiles and heavy-duty diesel trucks, Environ. Sci.Technol., 27, 636–651.

Rogge, W. F., L. M. Hildemann, M. A. Mazurek, G. R. Cass, and B. R. T.Simoneit (1993c), Sources of fine organic aerosol. 4. Particulate abrasionproducts from leaf surfaces of urban plants, Environ. Sci. Technol., 27,2700–2711.

Rogge, W. F., L. M. Hildemann, M. A. Mazurek, G. R. Cass, and B. R. T.Simoneit (1996), Mathematical modeling of atmospheric fine particle-associated primary organic compound concentrations, J. Geophys. Res.,101, 19,379–19,394.

Rogge, W. F., L. M. Hildemann, M. A. Mazurek, G. R. Cass, and B. R. T.Simoneit (1997), Sources of fine organic aerosol. 8. Boilers burning no. 2distillate fuel oil, Environ. Sci. Technol., 31, 2731–2732.

Rolph, G. D. (2003), Real-time Environmental Applications and DisplaysYstem (READY) Website, Air Resour. Lab., Natl. Oceanic Atmos. Ad-min., Silver Spring, Md. (Available at http://www.arl.noaa.gov/ready/hysplit4.html)

Schauer, J. J., W. F. Rogge, L. M. Hildemann, M. A. Mazurek, and G. R.Cass (1996), Source apportionment of airborne particulate matter usingorganic compounds as tracers, Atmos. Environ., 30, 3837–3855.

Schauer, J. J., M. J. Kleeman, G. R. Cass, and B. R. T. Simoneit (1999a),Measurement of emissions from air pollution sources. 1. C1 through C29organic compounds from meat charbroiling, Environ. Sci. Technol., 33,1566–1577.

Schauer, J. J., M. J. Kleeman, G. R. Cass, and B. R. T. Simoneit (1999b),Measurement of emissions from air pollution sources. 2. C1 through C30organic compounds from medium duty diesel trucks, Environ. Sci. Tech-nol., 33, 1578–1587.

Schauer, J. J., M. J. Kleeman, G. R. Cass, and B. R. T. Simoneit (2001),Measurement of emissions from air pollution sources. 3. C1 through C29organic compounds from fireplace combustion of wood, Environ. Sci.Technol., 35, 1716–1728.

Schauer, J. J., M. P. Fraser, G. R. Cass, and B. R. T. Simoneit (2002a),Source reconciliation of atmospheric gas-phase and particle-phase pollu-tants during a severe photochemical smog episode, Environ. Sci. Tech-nol., 36, 3806–3814.

Schauer, J. J., M. J. Kleeman, G. R. Cass, and B. R. T. Simoneit (2002b),Measurement of emissions from air pollution sources. 5. C1 through C32

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

13 of 14

D15303

organic compounds from gasoline-powered motor vehicles, Environ. Sci.Technol., 36, 1169–1180.

Sharma, D. N., A. A. Sawant, R. Uma, and D. R. Cocker III (2003),Preliminary chemical characterization of particle-phase organic com-pounds in New Delhi, India, Atmos. Environ., 37, 4317–4323.

Sheesley, R. J., J. J. Schauer, Z. Chowdhury, G. R. Cass, and B. R. T.Simoneit (2003), Characterization of organic aerosols emitted from thecombustion of biomass indigenous to South Asia, J. Geophys. Res.,108(D9), 4285, doi:10.1029/2002JD002981.

Simoneit, B. R. T. (1985), Application of molecular marker analysis tovehicular exhaust for source reconciliations, Int. J. Environ. Anal. Chem.,22, 203–233.

Simoneit, B. R. T. (1999), A review of biomarker compounds as sourceindicators and tracers for air pollution, Environ. Sci. Pollut. Res., 6, 159–169.

Simoneit, B. R. T., J. J. Schauer, C. G. Nolte, D. R. Oros, V. O. Elias, M. P.Fraser, W. F. Rogge, and G. R. Cass (1999), Levoglucosan, a tracer forcellulose in biomass burning and atmospheric particles, Atmos. Environ.,33, 173–182.

Smith, K. R. (1993), Fuel combustion, air pollution exposure, and health:The situation in developing countries, Annu. Rev. Energy Environ., 18,529–566.

United Nations Development Programme and World Bank Energy SectorManagement and Assistance Program (2004), Toward cleaner urban air inSouth Asia: Tackling transport pollution, understanding sources, ESMAPFormal Rep. 281/04, World Bank, Washington, D. C.

Venkataraman, C., G. Habib, G. A. Eiguren-Fernandez, A. H. Miguel, andS. K. Friedlander (2005), Residential biofuels in south Asia: Carbonac-eous aerosol emissions and climate impacts, Science, 307, 1454–1456.

Watson, J. G., N. F. Robinson, J. C. Chow, R. C. Henry, B. M. Kim, T. G.Pace, E. I. Meyer, and Q. Nguyen (1990), The USEPA/DRI chemicalmass balance receptor model, CMB 7.0, Environ. Software, 5, 38–49.

Watson, J. G., N. F. Robinson, E. M. Fujita, J. C. Chow, T. G. Pace,C. Lewis, and T. Coulter (1998), CMB8 applications and validation pro-tocol for PM2.5 and VOCs, Rep. no. 1808.2D1, Desert Res. Inst., Reno,Nv.

Yu, T.-S., T. W. Wong, X. R. Wang, H. Song, S. L. Wong, and J. L. Tang(2001), Adverse effects of low-level air pollution on the respiratory healthof school children in Hong Kong, J. Occupational Environ. Medicine, 43,310–316.

Zeng, Y., and P. K. Hopke (1989), Three-mode factor analysis: A newmultivariate method for analyzing spatial and temporal composition var-iation, in Receptor Models in Air Resources Management, edited by J. G.Watson, pp. 173–189, APCA Trans. no. 14, Air Pollut. Control Assoc.,Pittsburgh, Pa.

Zheng, M., G. R. Cass, J. J. Schauer, and E. S. Edgerton (2002), Sourceapportionment of PM2.5 in the southeastern United States using solvent-extractable organic compounds as tracers, Environ. Sci. Technol., 36,2361–2371.

Zheng, M., L. G. Salmon, J. J. Schauer, L. Zeng, C. S. Kiang, Y. Zhang, andG. R. Cass (2005), Seasonal trends in PM2.5 source contributions inBeijing, China, Atmos. Environ., 39, 3967–3976.

Zheng, M., G. S. W. Hagler, L. Ke, M. H. Bergin, F. Wang, P. K. K. Louie,L. Salmon, D. W. M. Sin, J. Z. Yu, and J. J. Schauer (2006), Compositionand sources of carbonaceous aerosols at three contrasting sites in HongKong, J. Geophys. Res., 111, D20313, doi:10.1029/2006JD007074.

�����������������������Z. Chowdhury, Environmental Health Sciences, School of Public Health,

University of California, Berkeley, 140 Warren Hall #7360, Berkeley, CA94720, USA. (zohir@berkeley.edu)A. G. Russell, Department of Civil and Environmental Engineering,

Georgia Institute of Technology, Atlanta, GA 30332, USA.L. G. Salmon, Environmental Engineering Sciences, California Institute

of Technology, Pasadena, CA 91125, USA.J. J. Schauer and R. J. Sheesley, Environmental Chemistry and

Technology Program, University of Wisconsin-Madison, Madison, WI53706, USA.M. Zheng, School of Earth and Atmospheric Sciences, Georgia Institute

of Technology, Atlanta, GA 30332, USA.

D15303 CHOWDHURY ET AL.: SOURCE APPORTIONMENT OF PM2.5 IN INDIA

14 of 14

D15303